Marcus Bointon marcus at
Wed May 16 13:06:46 UTC 2007

On 16 May 2007, at 13:46, Julio Leyva wrote:

> If I have two or three servers running memcached , how they  
> synchronized each other?

They are not - they have no idea other servers exist.

> How the application knows which memcached server to communicate to?

It's all done at the client end: every client must define an  
identical set of servers.

> So it means that any operation from then on will affect all the  
> servers in that list?
> memcache_delete(/* key */ TEXT, /* hold timer */ INTERVAL)
> memcache_delete(/* key */ TEXT)
> This command will affect both servers?

No, there is no redundancy. Only one server will get this command.  
This is what makes memcached scalable.

> What if one server goes down, when it comes back, how the get  
> synchronized?

They won't. It's just a cache, so requests for stuff on the missing  
server will fall through to the DB, eventually populating the cache  

Marcus Bointon
Synchromedia Limited: Creators of
UK resellers of info at hand CRM solutions
marcus at |

More information about the memcached mailing list